Personnel Groups
Currently there are four main groups at the LCB conducting their research in the areas of Computational Eukariotic Computational Genomics (Bongcam-Rudloff's group), Machine Learning in Bioinformatics (Komorowski's group) and Medical Bioinformatics (Gustafsson's group). The fourth group focused on Quantitative Genetics (Carlborg's group) is partially associated with LCB, partially with Swedish University of Agricultural Sciences. Please click on the tabs below to see more detailed description for each group provided along with research proposals from group members.
 Komorowski's
Carlborg's
Bongcam-Rudloff's
Gustafsson's
Group
Leader Jan Komorowski, Professor
Members Gunnar Andersson, Researcher
Robin Andersson, PhD Student
Susanne Bornelöv, PhD Student
Stefan Enroth, PhD Student
Marcin Kruczyk, PhD Student
Gunnar Andersson [top]
My research is focused on the use of machine learning for understanding and predicting disease. An important part is developing methods for feature selection and discretization. Recently I have initiated a collaborative project with Anders Wallin and Kaj Blennow at the department of neuroscience and physiology, at Göteborg University. The goal is to construct a rule-based model for predicting Alzheimer’s disease among patients with Mild Cognitive Disorder (MCI). I am using machine learning methods, including Random Forests and Rough Sets to analyze data from a major longitudinal study combining biochemical, genetic, neuroimaging and neuropsychological investigations. We will also try to identify novel markers from mass-spectrometry analyses of cerebrospinal fluid.

In a related project, I am investigating the possibilities of using similar methods to identify complex patterns in SNP data. Key problems include feature selection with dependant attributes and extraction of information from pedigrees.

* * *

Robin Andersson [top] [homepage]
Cancer progression often involves alterations in DNA copy number, indicating that failures in the mechanisms that maintain the integrity of the genome contribute to tumor evolution. The progression is enabled by the aberrant function of certain genes due to causes that may vary between tumors. Tumor suppressor genes may be inactivated by for example mutation or physical deletion. Likewise, oncogenes can be activated by mutation, structural rearrangement or amplification. Discovery and functional assessment of these genes is essential for understanding the biology of cancer and for clinical applications.

The development of genomic microarrays has enabled the genome-wide analysis of tumor cells. Genomic microarrays can be used in combination with comparative genomic hybridization (array-CGH) to assess DNA copy number imbalances in the genomes of many different tumor types. In a typical array-CGH measurement, total genomic DNA is extracted from test and reference cell populations, differently labeled with fluorescent dyes, and hybridized to genomic microarrays. The relative hybridization intensity of the test and reference signals at a given location is then proportional to the relative copy number of those sequences in the test and reference genomes. If the reference genome is normal, then increases and decreases in the intensity ratio indicate DNA copy-number variation in the genome of the test cells. In tiling genomic microarrays, the spots are often genomically overlapping clones that together cover the whole genome. The overlap of clones is used to guarantee full coverage and may enable the detection of chromosomal aberration breakpoints within spots.

The large amount of data generated from array-CGH experiments requires automatic procedures for the classification of copy number profiles of tumor DNA. Recently, various methods have been proposed for copy number profiling. The majority of these methods does not provide any statistical motivation for the generated profiles but relies solely on optimization criterions, convolutions, or ad hoc set thresholds. Moreover, we are not aware of any method that tackles the introduced dependency between clones when using tiling arrays.

My current research aims at developing statistical efficient methods for automatic copy number profiling from array-CGH data. Moreover, as current statistical methods for microarray data are based on cDNA analysis, methods for normalization and qualitative verification among others need to be revised dealing with array-CGH experiments.

Future projects will concern developing methods for classification based on copy number profiles, gene mining and the integration of different sources of data for machine learning.

My research is done in collaboration with the Molecular Oncology group led by prof. Jan Dumanski at the Department of Genetics and Pathology, Uppsala University.

Supervisor: Jan Komorowski Co-supervisor: Jan Dumanski

* * *

Stefan Enroth [top] [homepage]
My current research is focused partly on analysing Chromatin ImmunoPrecipitation (ChIP) tiling array data in a high throughput fashion with the immediate goals of identifying enriched regions for transcription factor bind sites and epigenic modification. The enriched genomic regions is then further analysed looking for consensus motifs or conserved sites. The long term goals include development of continuous shape recognition approaches to finding the enriched regions whilst increasing the precision of the predicted region. I am currently involved in two major projects in this area, one collaborating with Prof. Claes Wadelius’ group at the Rudbeck laboratory and through this project we have participated in an ENCODE project that aims at evaluating different approaches to ChIP-chip analysis. The results so far are very promising, the first project managed to pin point transcription factor binding sites at base pair resolution with the identification of known consensus motifs for the proteins investigated. The experimental work has been published and a method article is well on the way. In the ENCODE project our approach has stood up well to the techniques developed by major American labs and will most probably be included in the resulting publication of that analysis.

Another area of interest lies within development and implementation of hypothesis testing techniques in the annotation space of data generated from expression micro array or ChIP-chip experiments. The idea is not novel, comparing the number of annotation in a background distribution to a distribution in a selected set allows for statistical testing of significance, but there are still work needed to be done. In our case, the Gene Ontology is most often the source of annotation and since these are structured in a directed acyclic graph the terms are connected and the testing coupled. The aim is to develop algorithms that can summarise the results in a meaningful sense. The long term goals is to expand the knowledge that can be connected to the biological data using automated data mining in literature sources such as previously published articles. There is also a need to start taking higher order information into account. Genes are often co-regulated both in time and space and an extension of the existing tools into these domains is imminent.

* * *


© 2005. The Linnaeus Centre for Bioinformatics | webmaster